Use Simple Kmeans, DBScan, Hierachical clustering algorithms for clustering. Compare the performance of clusters by changing the parameters involved in the algorithms.

In [63]:
import pandas as pd
import numpy as np
HRTU_2=pd.read_csv("/content/drive/MyDrive/Datasets/HTRU_2.csv",header=None)
HRTU_2
Out[63]:
0 1 2 3 4 5 6 7 8
0 140.562500 55.683782 -0.234571 -0.699648 3.199833 19.110426 7.975532 74.242225 0
1 102.507812 58.882430 0.465318 -0.515088 1.677258 14.860146 10.576487 127.393580 0
2 103.015625 39.341649 0.323328 1.051164 3.121237 21.744669 7.735822 63.171909 0
3 136.750000 57.178449 -0.068415 -0.636238 3.642977 20.959280 6.896499 53.593661 0
4 88.726562 40.672225 0.600866 1.123492 1.178930 11.468720 14.269573 252.567306 0
... ... ... ... ... ... ... ... ... ...
17893 136.429688 59.847421 -0.187846 -0.738123 1.296823 12.166062 15.450260 285.931022 0
17894 122.554688 49.485605 0.127978 0.323061 16.409699 44.626893 2.945244 8.297092 0
17895 119.335938 59.935939 0.159363 -0.743025 21.430602 58.872000 2.499517 4.595173 0
17896 114.507812 53.902400 0.201161 -0.024789 1.946488 13.381731 10.007967 134.238910 0
17897 57.062500 85.797340 1.406391 0.089520 188.306020 64.712562 -1.597527 1.429475 0

17898 rows × 9 columns

In [64]:
df=HRTU_2.copy()
df=df.drop(columns=8)
from sklearn import preprocessing
rdf=pd.DataFrame(preprocessing.scale(df))
rdf
Out[64]:
0 1 2 3 4 5 6 7
0 1.149317 1.334832 -0.669570 -0.400459 -0.319440 -0.370625 -0.072798 -0.287438
1 -0.334168 1.802265 -0.011785 -0.370535 -0.371102 -0.588924 0.504427 0.211581
2 -0.314372 -1.053322 -0.145233 -0.116593 -0.322107 -0.235328 -0.125996 -0.391373
3 1.000694 1.553254 -0.513409 -0.390178 -0.304404 -0.275666 -0.312265 -0.481300
4 -0.871402 -0.858879 0.115609 -0.104866 -0.388010 -0.763111 1.324026 1.386794
... ... ... ... ... ... ... ... ...
17893 0.988208 1.943284 -0.625655 -0.406697 -0.384010 -0.727295 1.586054 1.700034
17894 0.447319 0.429062 -0.328831 -0.234643 0.128776 0.939926 -1.189159 -0.906574
17895 0.321842 1.956220 -0.299334 -0.407492 0.299137 1.671568 -1.288079 -0.941330
17896 0.133628 1.074510 -0.260050 -0.291041 -0.361967 -0.664857 0.378257 0.275850
17897 -2.105762 5.735470 0.872674 -0.272508 5.961291 1.971546 -2.197327 -0.971052

17898 rows × 8 columns

In [65]:
from sklearn.decomposition import PCA
pca=PCA(n_components=2)
ddf=pca.fit_transform(rdf)
ddf
Out[65]:
array([[-1.27884879, -1.27313333],
       [-1.02055347, -0.20116189],
       [ 0.18828859,  0.43211417],
       ...,
       [ 0.68183395, -2.6861702 ],
       [-1.10391246, -0.11650828],
       [ 3.71252289, -4.90385741]])
In [66]:
from sklearn.cluster import KMeans
import matplotlib.pyplot as plt
e=[]
for i in np.arange(2,10):
    clusters=KMeans(n_clusters=i)
    clusters.fit(ddf)
    e.append(clusters.inertia_)
ax=plt.figure(dpi=180).gca()
plt.plot(np.arange(2,10),e)
plt.xlabel("Clusters")
plt.ylabel("Squared sum of Error")
plt.show()
In [67]:
clusters= KMeans(n_clusters=3)
clusters.fit(ddf)
labels=clusters.labels_
ax=plt.figure(dpi=180).gca()
plt.scatter(ddf[:,0],ddf[:,1],c=labels,edgecolors='k',)
Out[67]:
<matplotlib.collections.PathCollection at 0x7f72ec284410>

Silouhette Assessment method

In [68]:
from sklearn.cluster import KMeans
import matplotlib.pyplot as plt
from sklearn.metrics import silhouette_score
ss=[]
for i in np.arange(2,10):
    clusters=KMeans(n_clusters=i)
    clusters.fit(ddf)
    ss.append(silhouette_score(ddf,clusters.labels_))
In [69]:
ax=plt.figure(dpi=180).gca()
plt.plot(np.arange(2,10),ss)
plt.xlabel("Clusters")
plt.ylabel("Silhouette score")
plt.show()
In [70]:
from sklearn import cluster, datasets, mixture
nsamples=1500
circle=datasets.make_circles(nsamples,noise=0.01)
moon=datasets.make_moons(nsamples,noise=0.01)
blob=datasets.make_blobs(nsamples)
random=np.random.rand(nsamples,2),None
In [71]:
DS=[circle,moon,blob,random]
SDS=[]
from sklearn import preprocessing
for i,j in DS:
    SDS.append(preprocessing.scale(i))

Silhouette Score for Toy datasets

In [72]:
k=0
f,axis=plt.subplots(1,4,figsize=(40,8))
for ds in SDS:
    ss=[]
    for i in np.arange(2,10):
        clusters=KMeans(n_clusters=i)
        clusters.fit(ds)
        ss.append(silhouette_score(ds,clusters.labels_))
    axis[k].plot(np.arange(2,10),ss)
    plt.xlabel("Clusters")
    plt.ylabel("Silhouette score")
    k=k+1
In [73]:
import numpy as np # linear algebra
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
from mpl_toolkits import mplot3d
import numpy as np
import matplotlib.pyplot as plt
In [74]:
from sklearn.cluster import AgglomerativeClustering
hei_clust = AgglomerativeClustering(n_clusters=3,linkage='average')
hei_clust.fit(ddf)
new_labels = hei_clust.labels_
plt.figure(figsize=(12,9))
plt.scatter(ddf[:,0],ddf[:,1],c = new_labels,cmap='autumn_r', edgecolors='K',s=50)
Out[74]:
<matplotlib.collections.PathCollection at 0x7f72eb283950>

Agglomerative on HRTU2

In [75]:
Agg=AgglomerativeClustering(n_clusters=3,linkage='single' )
H=Agg.fit(ddf)
new_labels=H.labels_
plt.figure(figsize=(12,9))
plt.scatter(ddf[:,0],ddf[:,1], c=new_labels,cmap='autumn_r', edgecolors='K',s=50)
Out[75]:
<matplotlib.collections.PathCollection at 0x7f72e57d8a10>
In [76]:
Agg=AgglomerativeClustering(n_clusters=3,linkage='complete' )
H=Agg.fit(ddf)
new_labels=H.labels_
plt.figure(figsize=(12,9))
plt.scatter(ddf[:,0],ddf[:,1], c=new_labels,cmap='autumn_r', edgecolors='K',s=50)
Out[76]:
<matplotlib.collections.PathCollection at 0x7f72e6af2e90>

DBSCAN on HRTU2

In [77]:
from sklearn.cluster import DBSCAN
db=DBSCAN(eps=0.8,metric="euclidean")
dbClust=db.fit(ddf)
new_labels=dbClust.labels_
plt.figure(figsize=(12,9))
plt.scatter(ddf[:,0],ddf[:,1], c=new_labels,cmap='jet', edgecolors='K',s=50)
Out[77]:
<matplotlib.collections.PathCollection at 0x7f72eb903150>

KMeans on Toy Datasets

In [78]:
for ds in SDS:
    f,axis=plt.subplots(1, 6,figsize=(28,4))
    for i in np.arange(2,8):
        clusters= KMeans(n_clusters=i)
        clusters.fit(ds)
        labels=clusters.labels_
        axis[i-2].scatter(ds[:,0],ds[:,1],c=labels,edgecolors='k',)
    plt.show()

Agglomerative clustering on Toy Datasets

In [79]:
for ds in SDS:
    f,axis=plt.subplots(1, 6,figsize=(28,4))
    for i in np.arange(2,8):
        hei_clust = AgglomerativeClustering(n_clusters=3,linkage='average')
        hei_clust.fit(ds)
        new_labels = hei_clust.labels_
        axis[i-2].scatter(ds[:,0],ds[:,1],c=new_labels,edgecolors='k',)
    plt.show()

DBSCAN on Toy Datasets

In [80]:
from sklearn.cluster import DBSCAN
for ds in SDS:
    f,axis=plt.subplots(1, 6,figsize=(28,4))
    for i in np.arange(2,8):
        db=DBSCAN(eps=(i-1)/10.0,metric="euclidean")
        clusters=db.fit(ds)
        new_labels=clusters.labels_
        axis[i-2].scatter(ds[:,0],ds[:,1],c=new_labels,edgecolors='k')
    plt.show()

Dendogram

In [81]:
import numpy as np
from sklearn import cluster, datasets, mixture
from itertools import cycle, islice
np.random.seed(0)

no_of_samples=100

noisy_circles,label1=datasets.make_circles(n_samples=no_of_samples, factor=0.5, noise=.05)
In [82]:
d=preprocessing.scale(noisy_circles)
In [83]:
from scipy.spatial import distance_matrix
dist=distance_matrix(d,d)
dist
Out[83]:
array([[0.        , 1.79890961, 3.49251197, ..., 2.0277973 , 3.32399774,
        3.10942345],
       [1.79890961, 0.        , 2.09664675, ..., 1.61429562, 1.62202173,
        1.3105377 ],
       [3.49251197, 2.09664675, 0.        , ..., 1.75454017, 0.93877394,
        1.67991193],
       ...,
       [2.0277973 , 1.61429562, 1.75454017, ..., 0.        , 2.07916566,
        2.36977863],
       [3.32399774, 1.62202173, 0.93877394, ..., 2.07916566, 0.        ,
        0.76744242],
       [3.10942345, 1.3105377 , 1.67991193, ..., 2.36977863, 0.76744242,
        0.        ]])
In [84]:
from scipy.cluster import hierarchy
Z=hierarchy.linkage(dist,'average')

/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:2: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix
In [85]:
dendro=hierarchy.dendrogram(Z,)
In [88]:
!cp  "/content/drive/MyDrive/Colab Notebooks/Practical 6.ipynb" ./
!jupyter nbconvert --to html "Practical 6"

[NbConvertApp] Converting notebook Practical 6.ipynb to html
[NbConvertApp] Writing 6897077 bytes to Practical 6.html